FR2929472A1 - METHOD FOR PROCESSING INTERFERENCE IN AN XDSL SYSTEM - Google Patents

Abstract

The invention relates to a method of processing interference affecting N signals respectively transmitted by N terminals belonging to a communication network, said N transmitted signals forming the components of a vector s, said method comprising at least one phase of estimating components of the vector s, comprising at least a first estimation step during which at least one component of the vector s is estimated as a function of at least one other component of the vector s estimated previously, said method further comprises: at least one step of changing a vector base in which the vector s is expressed, before at least a second step of estimating the components of the vector s; - a step of selecting at least one component of the vector s estimated at during one of said estimation steps; a step of reconstituting the vector s processed from said selected components.

Description

The invention relates to the technical field of transmissions of signals in channels of the same transmission system. In access networks, it is known to use data transmission techniques of the DSL type (for Digital Subscriber Line or digital subscriber line in French). Such techniques use a multi-carrier modulation DMT type (for Discrete MultiTone, or discrete multitonality in French). In a copper pair access network, each constituent active line of the cable connects a multiplexer, arranged for example at an access node, to an end user located downstream of the access node. For a set of end users given, the active lines can be grouped into a single cable and connected to the output of a first multiplexer. These active lines are managed by a first telecommunications operator. While other active lines of the same cable can be managed by at least one other telecommunication operator and be connected to another DSLAM multiplexer also located at the access node. In the following, the transmission system is the active lines managed by a given telecommunication operator inside a sheath and flowing to the same multiplexer. The signals of the same cable are all transmitted in a narrow frequency band. If an active line of a cable is considered, the signals transmitted by the other active lines of the cable cause interference generating a noise which disturbs the signals transmitted on the active line and reduces the transmission rate for this active line. Such noise is particularly difficult to eliminate when it is generated by the external channels of the transmission system, for which the system operator has no information. A method of handling interference in a DSL system is known. The method includes a first phase of bleaching external noises to the transmission system and a second phase of simultaneous processing of interferences internal to the system and interferences generated during the bleaching phase. The first phase comprises, for example, a step of Cholesky decomposition of a spatial noise correlation matrix Rn (noise external to the system plus white noise, it is considered that there are no internal noise in the system) present on the active lines of the transmission system considered, expressing the correlation between the different sources of noise. It also comprises a step of multiplying the signal received by the inverse of a matrix G resulting from the Cholesky decomposition of the noise spatial correlation matrix R.

The second phase then comprises a step of decomposing, according to the QR method, the product of the channel matrix by the inverse of the matrix G into a product of a triangular matrix R and an orthogonal matrix Q. The process comprises then a step of canceling the noise. With such a method, a value of the signal transmitted by the terminal which is designated Nth terminal, is estimated only according to the received signal, while a value of the signal transmitted by the terminal designated terminal Zef is estimated according to the signal received, and according to all the other signals already estimated (Nth, N_1th, 2nd). Such an estimate proves to be sufficient for the signal emitted by the terminal terminal, since it benefits from the estimates of the signals transmitted by the other terminals. The corresponding active line then has a satisfactory quality of service, particularly as regards the throughput. However, the inventors of the present patent application have found that the active lines associated with high-ranking terminals such as the terminal N do not have a sufficient quality of service. The invention mitigates this disadvantage by proposing an interference processing method affecting N signals respectively emitted by N terminals belonging to a communication network, said N transmitted signals forming the components of a vector s, said method comprising at least one phase method for estimating the components of the vector s, comprising at least a first estimation step during which at least one component of the vector s is estimated as a function of at least one other component of the vector s estimated previously. Such a method further comprises: at least one step of changing a vector base in which the vector s is expressed, prior to at least a second step of estimating the components of the vector s,

a step of selecting at least one component of the vector s estimated during one of said estimation steps;

a step of reconstituting the vector s processed from said selected components.

The vector s represents the signals emitted in at least two distinct vector bases, the vector s expressed in a first of the bases being linked by known relations to the vector s expressed in the second base. The estimation of the components of the vector s is then carried out in each of the vector bases. The components estimated least precisely in the first vector base are different from those estimated least precisely in the second vector base, when the two vector bases are chosen judiciously. Thus, we have several distinct estimates of the same vector s and we can choose, for each coordinate of the vector s, the best estimate.

The number of vector bases used for the estimation can obviously be modulated according to the parameters related to the transmission so that all the coordinates are estimated in a sufficiently precise way and, consequently, so that all the channels present a sufficient quality of service.

The invention therefore reconstitutes the vector s by improving the overall performance of the active lines of the transmission system, in particular by providing each terminal with a satisfactory bit rate for transmission.

In a particular embodiment, the second estimation step relates to a modified vector s in which one of said components has been replaced by said selected corresponding component.

This estimation is carried out according to the formula: Sk Yk ak, lSl ak k 1- which makes it possible to easily estimate the values of the various components of the vector s. In this formula, sk is the component corresponding to the signal emitted by the kth terminal of the system, @k is the signal received by the multiplexer and corresponding to the signal emitted by the kem terminal of the system and processed mathematically, and

at,,; are the coefficients of a given processing matrix.

Thus, it is possible to specify and improve the estimation of the components of the vector s. By replacing a component of the vector s with the component selected at the end of the first estimation step, a more accurate estimation of the components of the vector s is obtained. Advantageously, said vector bases are chosen according to parameters relating to a quality of service required for said transmitted signals.

In this way, it is avoided to determine vector bases automatically, if, for the quality of service required by a terminal, an estimate already made is sufficient. In a particular embodiment, a step of estimating said components comprises a first step of eliminating internal interference to the system affecting said components, and a second step of eliminating interference between the system and at least one external interference device. affecting said components. By external disturbance is meant not only terminals external to the system but also other types of disturbers, transmitting signals of a type other than those of the system (for example radio frequency signals) and interfering with the signals of the system. Thus, before carrying out the estimation of the vector s, we act on the vector s received to make sure that the share of the noise in the estimated vector s is as negligible as possible and that the estimation of the components of the vector s is as precise as possible. Optionally, the method comprises a step of Cholesky decomposition of a noise spatial correlation matrix of the received signals, the second processing step comprising a step of multiplication by the inverse of the matrix resulting from the Cholesky decomposition of the vector s received.

It is known that a spatial correlation matrix of noise expresses the statistical laws linking the noises coming from different sources, that is to say the influence that the different noises have on one another. A Cholesky decomposition of a given matrix M is the expression of this matrix according to the formula M = G x G "where G is an upper triangular matrix and G" is the transconjugate of the matrix G, that is, say the transpose of the matrix G in which all the complex numbers have been replaced by their conjugates. Such a treatment makes it possible to decorrelate the different sources of noise. Advantageously, the second processing step comprises a step of decomposing a matrix representing the response of the system into a product of an upper triangular matrix and an orthogonal matrix. These different decomposition techniques allow an easier estimation of the different components of the vector s.

Another object of the invention is an interference processing device affecting N signals respectively transmitted by N terminals belonging to a communication network, said N transmitted signals forming the components of a vector s, said device comprising means for estimation of the components of the vector s intended to implement at least a first estimation step during which at least one component of the vector s is estimated as a function of at least one other component of the vector s estimated previously. Such a device further comprises: means for changing a vector base in which the vector s is expressed, before at least a second step of estimating the components of the vector s, means for selecting at least one a component of the vector s estimated during one of said estimation steps; means for reconstituting the vector s processed from said selected components. The invention also relates to a computer program, characterized in that it comprises program code instructions for implementing the steps of the method described above when the program is executed by a processor. The reconstitution device is optionally integrated with a signal multiplexer, in particular a DSLAM multiplexer. Another object of the invention is a computer program capable of controlling the implementation of a method according to the invention. The invention will be better understood on reading the description which follows, given solely by way of example and with reference to the drawings in which: FIG. 1 represents a transmission system comprising a reconstitution device according to a mode FIG. 2 is a diagram showing a reconstitution method according to one embodiment of the invention; FIG. 3 is another diagram representing the reconstitution method of FIG. 2, FIG. 4 is a diagram showing a method according to another embodiment of the invention; FIG. 5 is another diagram showing the reconstitution method of FIG.

First Embodiment: FIG. 1 shows terminals T1 1 T2,..., TN 1, T1 2, T2 TM 1, N M formed for example by the modems of the users. Terminals T1 to TN are managed by a first telecommunications operator. Terminals T1., T2, and TM are managed by a second telecommunications operator distinct from the first operator. Each terminal T1,..., TN is connected to a DSLAM multiplexer 12 common to these terminals via respective active lines C1, C2, ..., CN. T1 terminals. to TM, are connected to another DSLAM multiplexer 12 'which is common to them, via active lines C1., C2', CM '. All the active lines C1, ..., CN, Cl ', ..., CM' are located in the same cable 14 and each correspond to a pair of copper wires constituting the cable 14. These active lines make it possible in particular to transmit DSL signals transmitted by each terminal to the multiplexer 12, 12 'concerned. The terminals T1,..., TN and the active lines C1,..., Associated CN and the multiplexer 12 form a first transmission system managed by the first operator. The DSLAM multiplexer 12 comprises a device 16 for processing interference affecting signals s1,..., SN, transmitted by the respective terminals T1,..., TN from the received signals y ,, ..., yN, by the multiplexer 12. This device 16 comprises a pre-programmed card integrated in the DSLAM comprising a memory 18 forming a storage means of a computer program for implementing a method described below, and means 20 for controlling the execution of the computer program, these means including a processor. We will now describe, with reference to Figures 2 and 3, a reconstitution method according to a first embodiment of the invention. This process follows the reception by the DSLAM multiplexer 12 of signals emitted by the terminals T1,..., TN, transmitted by the active lines C1,. The signals received by the multiplexer 12 are not similar to the signals emitted by the terminals T1,..., TN, since they have been disturbed during their transmission, in particular by signals z emitted by the terminals Tv,. TM, external to the system. This method makes it possible to reconstitute signals close to the signals emitted by the terminals T1,..., TN, so as to avoid any loss of flow in the active lines C1,..., CN.

The method comprises a first step E1 of reception by the DSLAM multiplexer 12 of the different signals transmitted by the terminals T1,..., TN. The received signals being expressed in the following way: y = Hs + Az + n (1) In this formula: - s is a vector representing the signals emitted by the different terminals belonging to the system, -7- - H is the matrix of system response, - A is a matrix representing the coupling of the signals emitted by terminals external to the system with the signals transmitted by the terminals of the system, - z is a vector representing the signals emitted by the external terminals T, T. 5 system and - n is a vector representing the noise inherent in the system. The expression of the received signal y in the formula (1) is frequency, that is to say that a received signal is expressed for each carrier frequency f of the transmitted signals. The method is described for a given carrier frequency, therefore, in order to simplify the notations, the index f is suppressed. Each of the N signals emitted by the terminals T T,..., TN is represented by a reference vector. Each of these reference vectors corresponds, therefore, to an active line of the system. The set of these N reference vectors constitutes a reference vector base BR in which the vector s representing the signals emitted by the different terminals T is expressed at TN and the vector y representing the signals received by the multiplexer DSLAM. Thus, when the vector s representing the signals emitted by the terminals T, at TN is expressed in the reference vector base BR, each of the N components of this vector s corresponds to the signal emitted by a terminal T. It is the same for the vector of received signals y, each coordinate of the vector y corresponds to a particular received signal and therefore to the active line of the corresponding system. It is considered that the DSLAM multiplexer 12 knows the spatial correlation matrix of the noise Rn., Present on the various active lines of the system. This matrix Rn., Has the following expression: Rn = AAH Z + ΔI The method comprises a step E2 of calculating the inverse G- 'of the matrix G resulting from the Cholesky decomposition of the matrix Rä • n. spatial correlation of noise. During a step E3, the vector of the received signals is multiplied by the matrix G. The vector of the processed signals is then expressed as follows: 30 = G-y = wy with: H = WH, H being called the equivalent system matrix and W being the inverse of the matrix G (6) -8- - ta = W (Az + n) represents a combination of Internet noise and external to the system The method comprises a step E4 of decomposition of the matrix H into a product of an upper triangular matrix R and a unit matrix Q, so that: H = QR = QRs + n An upper triangular matrix is a matrix in which all the coefficients situated above the diagonal of the matrix, as well as the diagonal coefficients are harmless. In an upper triangular matrix, the coefficients located below the diagonal of the matrix are null. The method then comprises a step E5 for multiplying the vector @ by the transconjugated matrix of the matrix Q: = QHÿ = Rs + Q "The method then comprises a step E6 of successive estimations of each signal s2i ..., sk, ..., SN emitted by the terminals Ti to TN At this step, an estimate of each coordinate of the vector s is calculated as follows: 1 N s = ù / tYiù ~, r ~, ksk) (7) Step E3 constitutes a step of whitening the internal and external noises to the system and the set PH1 of steps E4 to E6 allows the estimation of the signals emitted by each terminal, the method also comprises a step E7 of determining one or several vector bases in which the received signals will also be expressed, these vector bases are chosen as a function of QoS parameters requested by each terminal, these parameters notably including a parameter relating to the requested bit rate, depending on the quality of service. for example, the subscription subscribed, and / or a parameter relating to the state of the queue of each terminal, so as to make a better estimate of the signal transmitted by the terminal that has the most information to transmit, and / or a parameter relating to the state of each active line, so as to make a better estimate for an active line having the highest attenuation or undergoing the most the impact of the external noise. In a particular embodiment of the method that is the subject of the invention, the device 16 applies, during the step E7, a permutation function P i, determined according to the parameters listed above, to the vectors of the base of reference BR in order to obtain one or more bases B0) in which, for a given vector X, expressed in the base BR, the vector X; corresponding, expressed in the base B ° '), is such that X; = P; xX. In each base B ('), the reference vectors of the base BR do not change, only the order of these vectors in the family formed by the base is changed. For a matrix M expressed in the reference base BR (for example, the matrices H or Vl /), the corresponding matrix M; expressed in the base B °) will be: M; = P; xMxP; T 10 because of the classical formulas and the properties of the permutation matrices. In the following, it is considered that N = 6, that all terminals T ,, ... TN require identical quality of service and that the estimates in the reference base make it possible to obtain a sufficient quality of service only for the signals issued by two terminals, T, and T2. In order for the qualities of service to be in conformity with those requested, two distinct bases of the base BR, the bases B 'and B ", are determined in which the transmitted signals are expressed as follows: / s',' 1 / s3 \ / 0 0 1 0 0 0 0 s, 0 0 0 1 0 0 s'3 s5 0 0 0 0 1 0 s' = which means that P '= and s'4 s, 0 0 0 0 0 0 1 s' 5 s, 1 0 0 0 0 0 \ S'6 / Q, \ s2 / e, v0 1 0 0 0 0 / / s ",` s5 0 0 0 0 1 0 \ s "2 s6 0 0 0 0 0 1 s "1 0 0 0 0 0 s" = s 3 = s 2 which means that p "= 0 1 0 0 0 0 0 ~ as" 5 s3 0 0 1 0 0 0 s, / B 0 0 0 1 The method then comprises an expression step E8 in the vector base B 'and in the vector base 8 ", using the vector base change matrices determined in step E7, of the received signals. During a step E9, the set of steps PH1 for estimating the transmitted signals is then reproduced in each of the vector bases B 'and B ".The set of steps E1 to E9 constitutes one embodiment of the In this embodiment, the estimation of each coordinate of the vector s represents the transmitted signals expressed in a first vector base, for example the vector base B '. , is done independently of the estimation of the same components of the vector s representing the transmitted signals expressed in a second vector base, for example the vector base B ". We then obtain two estimates of the vector s representing the signals emitted in two distinct vector bases. Following these steps E1 to E9, the method according to the invention comprises, for each vector base, a step E10 for selecting the estimated components, or estimated signals, of the vector s representing the transmitted signals. Here the estimated components are those whose index in the vector base considered are the smallest. For example, the estimated signals for which a satisfactory estimate is obtained are, in the base B ', the estimated signals and s 2 (corresponding to the estimated signals s 3 and s 4 of the reference base BR) which have been estimated according to estimates of the signals s'6 S-5, s 4. 3 for the signal S''3 and S''6, s 5 s 4, 3, s 2 for the signal In the base B "the estimated signals for which a satisfactory estimate is obtained are the estimated signals s", and s "2 (corresponding to the signals s5 and s6 of the reference base BR.) In this step, therefore, the vector of the following estimated transmitted signals expressed in the reference base is obtained: s = sl s2 S'3. With the method according to the invention, it is therefore possible to obtain satisfactory estimates for all the signals transmitted, Another embodiment of the method which is the subject of the invention, referred to as series mode, will now be The method comprises a first step F1 for receiving, by the DSLAM multiplexer 12, the different signals transmitted by the terminals T 1,..., TN, the received signals being expressed as follows: y = Hs + Az + n (1) The DSLAM multiplexer 12 is considered to know the spatial correlation matrix This method comprises: a step F 2 of Cholesky decomposition of the spatial noise correlation matrix of received signals, followed by a step F3 for multiplying the vector of the received signals y the matrix G- '. The vector of the processed signals is then expressed as follows: ÿ = G-1y = Wy = Hs + (6) with: - H = WH H being called the equivalent system matrix and W being the inverse of the matrix G - fi = W (Az + n) represents noise a combination of internet noises and external to the system. The method then comprises a step F4 for decomposing the matrix H into a product of an upper triangular matrix R and a unit matrix Q, so that: H = QR @ = QRs + n The method then comprises a step F5 multiplying the signals ÿ by the transconjugated matrix of the matrix Q: = QHÿ = Rs + Q''n The method then comprises a step F6 of successive estimates of each signal s ,, s2, ..., sk, .. ., sN issued by terminals T1 to TN. During this step, an estimate of each coordinate of the vector s is calculated as follows: k = i + 1 The method also comprises a step F7 for determining one or more vector bases, in which the received signals will also be expressed y from one or more criteria mentioned above. The method then comprises an expression step F8 in the vector base B ', using the vector base change matrix determined in step F7, the received signals y. This step is similar to step E8. 1 N = ù (Y; û r; kSk) (7) r; As in the first embodiment described above, it is considered that N = 6, that all the terminals Ti,... TN demand identical quality of service and that the estimates in the reference base make it possible to obtain a sufficient quality of service only for the signals emitted by two terminals, T, and

T2.

Thus, the estimation of the signals in the reference base BR is satisfactory for the signals S1 and S2. The method then comprises a step F9 of successive estimates of each signal s ,, s2i ..., Sk, .. ,, SN emitted by the terminals T, to TN in the base B '. In this step, an estimate of each coordinate of the vector s is computed as follows: ## EQU1 ## P (P) = i; (P) ù r (, k) Sk (P) rik = r + I '"' (P) ^ (P- (1) with Sk s ~ 'P (i ()) where: - S Î P) is the estimated value of the signal associated with the ith coordinate of the vector s in the base B (P) issue of the permutation permutation of the reference base (for example B (') = B'), (P) y, is the ith coordinate of the vector associated with the signal received and processed in the base B ('), - ri (f ) is the coordinate associated with the ith line and the kth column of the matrix R in the base B (P)

- K is the number of terminals for which each permutation is considered satisfactory (in the previous example, K = 2), and - j) is the value of j permuted with the help of the permutation pe (in the given example above S5 = Spi l`rô (5)) = S (D (5) = SI). For example, in the vector base B 'the components s' and' 6 are replaced by the estimated and corresponding components. Indeed, s'5 = s, and s'6 = s2 in view of the relations between the bases BR and B '. It is not necessary to estimate these two components.

Thus, the other signals are estimated in the base B 'as a function of signal values estimated in the base BR according to the formula: In this context, the estimation of the signals is more precise and the estimation of the signals s',, s 3 corresponding to the terminals T3, T4 and T5 is sufficient for these terminals requiring a quality of service identical to that of the terminals T, and T2. In addition, in this example, the terminal T6 requires a lower quality of service than those of the other terminals and the estimate in the base B 'also proves to be sufficient for the signal corresponding to the signal s6 emitted by the terminal T6 . The step F9 described above is applied successively to all the selected permutation bases until a sufficient estimate of all the active lines of the system is obtained. As successive estimations are made, components of the vector s representing the transmitted signals expressed in a given vector base are replaced by estimates of these same components obtained during a previous estimation. This gives a satisfactory estimate for all the components of the vector s representing the transmitted signals. The method then comprises a step F10 for selecting the components of the vector s representing the estimated transmitted signals. The vector of the transmitted signals estimated in the reference base BR is expressed, for the example cited, during this step as follows: ## EQU1 ## It will be appreciated that the invention is not limited to the previously described embodiments. As a variant, the replacement of the estimated components of the transmitted signals expressed in the base B 'by the estimated components of the transmitted signals expressed in the base B is optional. In addition, the method is adjustable according to the number of terminals of the system and the number or nature of the parameters relating to the quality of service required by each terminal. For example, if the baseline estimate is sufficient for a single terminal, referred to as the first terminal, and all terminals have identical quality of service parameters, it is preferable to determine as many vector bases as there are terminals. Likewise, the criteria for determining the parameters to be considered for determining the permutation (s) to be carried out may vary. In the second embodiment, considering the state parameter of an active line, the most noise-sensitive active lines are determined by calculating the sum of the square modules of the terms of each column of the equivalent matrix H of the system. , the weakest terms corresponding to the active lines most sensitive to noise. In addition, the treatments applied to the vector y representing the signals received are not limited to the treatments described. Other known treatments can be applied to the vector for the purpose of reducing the influence of interference between the system and the external disturbers. Such processing may for example include a step of inverting the square root matrix of the noise spatial correlation matrix. Similarly, the treatment for the elimination of internal interference is not limited to the described treatments.

In addition, the chosen vector bases may be linked by relations other than permutation relations. Finally, the system is not limited to a copper wire pair cable.

Second Embodiment: Another Embodiment of the Invention Will Now Be Described This embodiment of the method that is the subject of the invention differs from those previously described in the steps of bleaching and estimation of the transmitted signals. In this embodiment, the method comprises a step G1 of multiplying the vector of signals received y by the inverse matrix of the response matrix of the system, so that the vector of the signals received and processed after step G1 is expressed as follows: 2929472 -15- = = s + n (2) where n = H = H - '(Az + n) This step corresponds to a step of processing the received signal in order to eliminate the internal interferences with system. It is considered that the DSLAM multiplexer 12 knows the spatial correlation matrix of the noise Rn ', n'. This matrix Rn. Has the following expression: ## STR1 ## In this formula: ## STR1 ## is the transconjugate of A, that is to say the transpose of this matrix in which the complex numbers have been replaced by their conjugates, - / is the identity matrix, and and are the respective powers of the external z signals to the system and the white noise n, and the correlation matrix of the yin noise,! a for expression 20 The method comprises Next, a Cholesky decomposition step G2 of the noise spatial correlation matrix In this step, the device 16 decomposes the spatial correlation matrix of the noise into a product of two matrices according to the formula: R; n = GxG "in where G is an upper triangular matrix and G "is transconjugated, this decomposition is carried out in a conventional manner, the method then comprises a step G3, in which a filter is applied to the vector ÿ representing the signals processed during the step G2 by multiplying this vector by a matrix W = G '. Thus, 30 ÿ = Wÿ = Ws + Wh '(4) This operation makes it possible to decorrelate the combination of the internal and external noises to the system and corresponds to a processing step of the processed signals in order to reduce -16-

the influence of the external disturbers on the active lines of the system, the external disturbers corresponding in particular to the terminals T'1, ..., T'N.

The method then comprises a step G4 of estimating each signal s ,, s2,..., Sk,..., SN emitted by the terminals T, to TN from the previous expression. Since the matrix W is the inverse of an upper triangular matrix, W is an upper triangular matrix. Hence: 1 N (Yi ù Wi, k sk) (5) wi, i kH-1 Thus, a value of the signal emitted by each terminal of the system T ,, ..., TN can be estimated as a function of a value of the corresponding received signal processed mathematically and, possibly, as a function of the estimated values of the signals transmitted by other terminals of the system.

Step G1 makes it possible to cancel the interferences internal to the system, the steps G2 to G3 form a set PH2 of steps of whitening of the noise, and the step G4 forming a step of estimating the emitted signals.

In this embodiment, the method which is the subject of the invention makes it possible to reduce the number and the complexity of the calculations to be performed in order to estimate the transmitted signals. Indeed, such a method requires fewer mathematical operations to obtain an estimate of the transmitted signals than the solution of the state of the art.

As a variant, the first processing step is not limited to a multiplication of the vector of the signals received by the inverse H "1 of the system response matrix, for example, it is still possible to simplify this step by decomposing the response matrix. of the system H in a sum of two matrices, so that:

H = D + E

with: - D including the diagonal terms of H and null non-diagonal terms, E including the non diagonal terms of H and null diagonal terms. From where :

H = D (I-D-1 E)

D "'E comprises terms smaller than 1, the matrix H having, because of its classical properties, diagonal order-of-magnitude terms higher than the non-diagonal terms, so that it can be approximated by following way using a series decomposition: H- 'pH = (I -D lE- (D -EE) 2) D-1 si = -17- The matrix D being a diagonal matrix, it is Even more simple to calculate U 'than F', this variant makes it possible to further improve the processing time of the estimation, since the step of estimating the signals emitted is a successive estimate. by applying the series of steps F8 to F9.

Claims (7)

REVENDICATIONS1. A method of processing interference affecting N signals respectively transmitted by N terminals belonging to a communication network, said N transmitted signals forming the components of a vector s, said method comprising at least one phase of estimating the components of the vector s, comprising at least a first estimation step in which at least one component of the vector s is estimated as a function of at least one other component of the vector s estimated previously, characterized in that said method further comprises: - at least a step of changing a vector base in which the vector s is expressed, before at least a second step of estimating the components of the vector s, - a step of selecting at least one component of the vector s estimated during one of said estimation steps; a step of reconstituting the vector s processed from said selected components. 2. Method according to claim 1, wherein said second estimation step relates to a modified vector s in which one of said components has been replaced by said selected corresponding component. The method of claim 1, wherein said vector bases are selected according to parameters relating to a quality of service requested for said transmitted signals. The method of claim 1, wherein a step of estimating said components comprises a first step of eliminating internal interference to the system affecting said components, and a second step of eliminating interference between the system and at least one an external disturbance affecting said components. 5. An interference processing device affecting N signals respectively emitted by N terminals belonging to a communication network, said N transmitted signals forming the components of a vector s, said device comprising means for estimating the components of the vector s intended to to implement at least a first estimation step during which at least one component of the vector s is estimated as a function of at least one other component of the vector s estimated previously, characterized in that said device further comprises: means for changing a vector base in which the vector s is expressed, prior to at least a second step of estimating the components of the vector s, means for selecting at least one component of the vector s estimated at during one of said estimation steps; means for reconstituting the vector s processed from said selected components. Computer program, characterized in that it comprises program code instructions for implementing the steps of the method according to claim 1 when the program is executed by a processor. 7. Data carrier, characterized in that it comprises storage means (18) of a computer program according to the preceding claim.